[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Inspiron 8000 IDE bus speed



On 15 Jan 2002, Simon Wong wrote:
> On Tue, 2002-01-15 at 10:42, Daniel Pittman wrote:
>> On 15 Jan 2002, Simon Wong wrote:
>> > I have noticed on booting that I have a message saying that an IDE
>> > bus speed of 33MHz is being assumed.
>> 
>> That's 33MHz *PCI* bus speed. Even if it says IDE everywhere near it.
> 
>     Uniform Multi-Platform E-IDE driver Revision: 6.31
>     ide: Assuming 33MHz system bus speed for PIO modes; override with
>     idebus=xx

yeah. The "system bus speed" it mentions is the system *PCI* bus. I
think it's stupid, too, but there doesn't seem to be much inclination to
change it.

Maybe Marcello, the 2.4 maintainer, would take a cosmetic patch if you
sent one. :)

>> > As far as I know the i8000 has a 66MHz bus speed.
>> 
>> That's unlikely. It may support UDMA-66 mode for the drives.
> 
> Sorry, I seem to have that mixed up.  It does support UDMA-66 for the
> IDE drives.

*nod*  It's fine; it's one of the more confusing details that tend to
get mixed around by the manufacturers. It's ever worse because it's so
little to do with real performance, too. :)

> I'll look at the doc on hdparm further to see if it's worth trying
> -X66 to get some more speed.

Given that you are using a laptop harddrive[1], you are /very/ unlikely
to see any real performance difference.

Performance on a single hard disk is dominated by seek time for
everything except streaming media.

Running the data transfer faster, such as UDMA-66 vs UDMA-33, generally
doesn't make a real difference to any situation using IDE disks because
one disk can't saturate the IO bus and two disks still lock each other
out, competing for the bandwidth.

So, it really doesn't make much difference. If you are using a UDMA, or
even a DMA, mode, there is little gain to be made by pushing it harder.
There is a (miniscule) theoretical gain using DMA over PIO, but it's not
really measurable.

Enabling 32-bit host transfers and unmasking of the IRQ with hdparm,
when it works[2], can make a difference, though -- especially the later
of those two.

        Daniel

Footnotes: 
[1]  It's a safe assumption, on this list, I figure.

[2]  Take backups, because it can corrupt data.

-- 
If you cannot get rid of the family skeleton, you may as well make it dance.
        -- George Bernard Shaw



Reply to: